Articles techniques et présentations

Endometriosis is a chronic inflammatory disease affecting roughly one in ten women, with no current cure, and existing therapies rely on systemic hormone suppressants that require daily compliance and produce significant side effects, creating a clear need for sustained-release alternatives. Our goal was to design Endostella, a subdermal implant that delivers leuprolide acetate—a GnRH agonist—at a controlled rate, with the specific design target of sustaining a therapeutic release of approximately 0.12 µg/day over a full 365 days from a single implant. Because the release profile had to be understood before any device was built, COMSOL Multiphysics was central to the design process from the start. We first modeled the implant as a diffusion-controlled reservoir system: a compressed drug core surrounded by a polyurethane membrane that serves as the rate-limiting barrier. Using the Transport of Diluted Species interface, we built a 3D model of the implant, defining the core as a high-concentration reservoir, the membrane as the controlling diffusion domain, and the surrounding tissue as a sink boundary; transport was governed by Fickian diffusion combined with a first-order term representing capillary clearance of the drug into the bloodstream, and a time-dependent study resolved the concentration gradient across the membrane and the cumulative drug released over a one-year horizon. Next, we calibrated the model against experiment, comparing its predicted release to Franz diffusion cell measurements and refining the effective diffusion coefficient until the model reproduced the observed release behavior, grounding the key transport parameter in data rather than estimation. With the calibrated model in hand, we performed a sensitivity analysis, sweeping membrane thickness, core dimensions, and drug loading across physically plausible ranges to map how each shaped the release curve, which identified the geometry—a 24 mm implant with a 2 mm core and a 0.24 mm membrane—that achieved the target release rate sustained across the full 365-day window. The calibrated model predicted a steady, near-zero-order release over the one-year period; this behavior follows directly from the reservoir architecture, because as long as the drug core remains saturated, the concentration at the inner membrane surface stays constant, holding the diffusion gradient—and therefore the release rate—nearly constant over time, and release rate and duration shifted predictably with membrane thickness and geometry alone, showing the design is tunable rather than fixed. Together, these results demonstrated that year-long, diffusion-controlled subdermal delivery of leuprolide acetate is feasible and tunable, and allowed us to converge on a viable implant geometry through simulation rather than repeated fabrication; on the strength of the model, we proceeded to manufacture the implant. Beyond its technical feasibility, the simulation-driven design yielded a low-cost intervention, with an estimated annual cost far below existing therapies such as Lupron Depot and Orilissa, and the same framework applies to any reservoir-membrane implant, offering a scalable, simulation-driven strategy for designing long-acting drug delivery systems in women's health and beyond.